1. Field of the Invention
The present invention is directed to a method and system for utilizing a shaped orifice or nozzle in a plasma processing system.
2. Discussion of the Background
During the fabrication of integrated circuits (IC), a conventional approach to oxide etch employs a capacitively coupled plasma (CCP) source, wherein a process gas comprising argon, CxFy (e.g. C4F8), and O2 is introduced to a low pressure environment to form plasma. Therefrom, the plasma dissociation chemistry is tuned for optimal production of the chemical reactant suitable for chemically reacting with the substrate surface material to be etched (i.e., CF2 for selective oxide etch). Moreover, the plasma further produces a population of positively charged ions (e.g. singly charged argon Ar+) suitable for providing energy to the substrate surface to activate the etch chemistry. In general, a substrate RF bias is employed to attract ions to the substrate surface in a controllable, directional manner to affect the ion energy at the substrate surface and to provide an anisotropic etch for desired feature side-wall profiles.
Due to the differing roles of the atomic, molecular and ionic species present in the plasma, it is believed that oxide etch comprises two fundamentally unique processes. Firstly, electrons are heated in the plasma whereby collisions with fluorocarbon species leads to dissociation and formation of radical species, e.g. CF3, CF2, CF, F, etc. And secondly, electrons are heated to energies sufficient to ionize argon atoms, whereby the resultant ions are utilized to energize substrate surface CFx/SiO2 chemical reactions.
For instance, referring to FIG. 1, an exploded view of an etch feature in an oxide layer is shown. In the plasma, fluorocarbon radicals are formed. Thereafter, they diffuse to the substrate and deposit onto the etch feature surfaces. Preferably, an increased concentration of CF2 radical local to the wafer surface can lead to several advantages (Nakagawa et al. 1998, Booth 1998, Kiss et al. 1992, Butterbaugh et al. 1991, Tatsumi et al. 1998), in particular: (1) the formation of a CFx polymer layer atop the patterned photoresist tends to protect the resist during the etch process for improved selectivity of SiO2-to-resist etch, (2) the formation of CFx polymer layer along sidewalls provides protection for improved etch anisotropy, and (3) the formation of CF2 at the bottom of a feature provides a suitable etch reactant for selective etch of oxide relative to silicon that produces volatile products, i.e. one of many chemical reactions can be 2CF2+SiO2→SiF4+2CO. Thereafter, the directional nature of the ion bombardment of the substrate surface, as shown in FIG. 2, leads to an anisotropic etch, wherein the argon ion energy is sufficient to activate the etch chemistry in the etch features.
One technique proposed in the archival literature as described above to improve high aspect ratio contact etch in oxide (e.g., etch rate, side-wall profile, selectivity, etc.) suggests optimization of the plasma chemistry to form CF2 radical. In doing so, studies have shown that the concentration of fluorocarbon radicals, particularly CF2, correlate well with τneσν, where τ is the gas residence time, ne is the electron density, σ is the dissociation collision cross-section, v is the electron velocity and σν is the integration of the product σν with the normalized electron energy distribution function (Tatsumi et al. 1998). Hence, conventional practice entails adjusting the plasma density to optimize the concentration of the preferred etch radical to achieve the enumerated conditions above and, in general, for oxide etch it can lead to limitations on the maximum etch rate. This shortcoming is often governed by the demand for meeting an etch selectivity specification or a side-wall profile specification. For instance, the etch rate is typically proportional to the plasma density (ion density equals the electron density for a quasi-neutral plasma, and either can be referred generally as the plasma density), whereas the etch selectivity can be inversely proportional to the plasma density once the plasma density is sufficiently large to produce a highly dissociated radical concentration (i.e., high fluorine radical concentration for etching oxide with CxFy process chemistry). Moreover, inappropriate chemistry due to excessively high dissociation rates can lead to inadequate protection of feature side-walls and therefore jeopardize side-wall profiles leading to an isotropic etch. Lastly, insufficient plasma density and low dissociation (i.e., high concentrations of CF3, CF2, etc.) can lead to etch stop due to material (i.e., C) build-up at the bottom of etch features. Therefore, due to the close relationship between the plasma density and the preferred radical concentration, this results in a very narrow parameter space wherein one must work to achieve marginally acceptable performance specifications for etch rate, etch selectivity and side-wall profile (or anisotropy). This is a major shortcoming for conventional hardware and process practice particularly since the etch requirements vary during the period one etches a deep, high aspect ratio contact.
For a typical shower-head gas injection system utilized in a conventional semiconductor processing device, the inject plate generally comprises an array of several hundred (several hundred to several thousand) inject orifices through which gas is introduced to the processing region at a flow rate equivalent to 100–1000 sccm argon. Furthermore, the injection orifice is typically a cylindrical orifice as shown in FIG. 3 characterized by a length L and diameter d, wherein the ratio of the orifice length to the orifice diameter L/d is greater than 10 (i.e., L/d>>1). For instance, a typical orifice diameter is 0.5 mm and a typical orifice length is 1 cm, and therefore the aspect ratio L/d=20.
As a consequence of this design, the gas effuses from the orifice exit with a very broad angular distribution characteristic of a low discharge coefficient orifice. The discharge coefficient of an orifice CD is given by the ratio of the real mass flow rate to the isentropic mass flow rate. The isentropic mass flow rate can be derived from the Euler equations (or inviscid Navier-Stoke's equations) for a quasi-one dimensional frictionless and adiabatic flow, viz.
                                          m            isentropic                    =                                                                      P                  t                                ⁡                                  (                                                            γ                      +                      1                                        2                                    )                                                                              -                  1                                /                                  (                                      γ                    +                    1                                    )                                                      ⁢                                                            2                  ⁢                  γ                                                  γ                  ⁢                                                                          ⁢                  R                  ⁢                                                                          ⁢                                                            T                      t                                        ⁡                                          (                                              γ                        +                        1                                            )                                                                                            ⁢            A                          ,                            (        1        )            where γ is the ratio of specific heats for the gas, R is the gas constant, Pt is the total pressure, Tt is the total temperature and A is the minimum cross-sectional (throat) area (i.e., A=πd2/4). When CD<<1, the total pressure recovery through the orifice is severely reduced and hence the angular distribution of the orifice flux becomes very broad. Therefore, a shortcoming of conventional gas injection system designs is a relatively low gas injection orifice discharge coefficient.
Furthermore, conventional systems suffer from a lack of control of the gas injection orifice discharge coefficient. In many cases, the gas injection orifice is subjected to erosion and, hence, the gas injection properties vary in time during a process, from substrate-to-substrate and batch-to-batch. Conventional systems do not monitor the state of a “consumable” gas injection system nor do they attempt to control the gas injection properties to prolong the life of a “consumable” gas injection system.
In addition to gas injection orifices with an uncontrollable discharge coefficient, conventional designs suffer from an additional shortcoming. That is the gas injection orifices are not oriented relative to one another to provide a uniform, directional flow local to the substrate surface.